Cumulative Network Meta-Analysis and Clinical Practice Guidelines - A Case Study on First-Line Medical Therapies for Primary Open-Angle Glaucoma

by Benjamin Rouse

A thesis submitted to the Department of Epidemiology of Johns Hopkins Bloomberg School of Public Health in conformity with the requirements for the degree of Master in Health Science, Epidemiology, Clinical Trials and Evidence Synthesis

Baltimore, Maryland May 2015

Abstract

Background

Clinical practice guidelines are statements of recommendations for patient care. Studies have shown that guideline recommendations do not always depend on evidence from clinical trials or systematic reviews. It is unknown whether no high quality evidence exists, evidence exists but authors were unaware of it, or advanced statistical methods were not available to them to address their questions. Our objective was to compare the guideline recommendations for first-line medical therapy for primary open-angle glaucoma (POAG) from each major update of the American Academy of

Ophthalmology’s (AAO) Preferred Practice Patterns (PPPs) with the actual evidence base available at the time.

Methods

We identified and extracted recommendations relevant to first-line medical therapy for

POAG from each version of the AAO PPP. We searched MEDLINE, EMBASE, and

CENTRAL for randomized controlled trials published up to March 2014. We analyzed intraocular pressure (IOP) outcome data as available at the time of each major guideline update. We used network meta-analysis to determine which of all drugs “works best.”

Results

We identified 9 versions of AAO’s guideline for POAG published between 1989 and

2010. Based on similarity in treatment recommendations or discussion, we grouped these

ii

guidelines into 5 sets: 1989-1992, 1996, 2000-2003; 2005-2006, and 2010. The 2010 guideline recommended prostaglandins as initial treatment, but previous sets presented treatment options without recommending one drug (or class) over another. Based on a series of network meta-analyses of trials published up to around the time of the latest guideline in each set, all drugs are more effective than placebo or no treatment at each time point, but effect size appears to decrease over time. Network meta-analysis indicated that the most effective drug and class (at time point analyzed) were: levobunolol and beta blockers (1991), levobunolol and alpha agonists (1995), travoprost and prostaglandins

(2002), bimatoprost and prostaglandins (2004 and 2009).

Conclusions

Network meta-analysis improves our understanding of the comparative effectiveness of multiple interventions. Had network meta-analysis been available, the AAO POAG PPP could have recommended prostaglandins (current first-line treatment) seven years before it actually did. Guideline developers should consider using results from network meta- analyses in forming future recommendations.

Advisor: Tianjing Li, PhD, MD, MHS

Second reader: Roberta W. Scherer, PhD, MS

iii

Acknowledgements

I would like to thank Tianjing Li for serving as my main thesis advisor. I would also like to thank Roberta Scherer for her assistance as a second reader and my academic advisor. I would like to thank Hwanhee Hong, Qiyuan Shi, Lijuan Zeng, Gillian

Gresham, Kexin Jin, Cesar Ugarte-Gil, Sueko Ng, and Nan Guo for their help with study screening, data abstraction, and data analysis. Finally, I would like to thank my family and friends for their encouragement throughout the program.

The project was funded by Grant 1 RC1 EY020140 and Grant 1 U01 EY020522,

National Eye Institute, National Institutes of Health, United States (PI: Dr. Kay

Dickersin).

iv

Table of Contents

1 Introduction ...... 1

2 Objective ...... 4

3 Methods...... 5

3.1 Guideline identification and extraction ...... 5

3.2 Systematic review and network meta-analysis...... 5

3.2.1 Eligibility criteria ...... 6

3.2.2 Identification of included studies ...... 6

3.2.3 Data abstraction and management ...... 7

3.2.4 Qualitative synthesis ...... 7

3.2.5 Quantitative synthesis ...... 8

3.2.6 Evaluation of network meta-analysis assumptions ...... 8

3.2.7 Measures of relative treatment effect ...... 9

3.3 Guideline and network meta-analysis comparison...... 10

4 Results ...... 11

4.1 Guideline identification and extraction ...... 11

4.2 Network meta-analysis ...... 11

4.2.1 Search results and general study characteristics ...... 11

4.2.2 Risk of bias ...... 13

4.2.3 Interventions ...... 13

4.2.4 Network meta-analysis outcomes ...... 14

4.2.5 Inconsistency...... 15

4.3 Guideline and network meta-analysis comparison...... 15

5 Discussion ...... 17

6 References ...... 22

7 Tables ...... 26

7.1 Table 1 Recommendations from AAO POAG PPPs ...... 26

v

7.2 Table 2 Summary estimates for intraocular pressure (mmHg) at 3 months derived from network meta-analysis ...... 27

7.2.1 Table 2a Network meta-analysis IOP estimates for drugs from studies published by 1991 ...... 27

7.2.2 Table 2b Network meta-analysis IOP estimates for drugs from studies published by 1995 ...... 28

7.2.3 Table 2c Network meta-analysis IOP estimates for drugs from studies published by 2002 ...... 29

7.2.4 Table 2d Network meta-analysis IOP estimates for drugs from studies published by 2004 ...... 30

7.2.5 Table 2e Network meta-analysis IOP estimates for drugs from studies published by 2009 ...... 31

7.2.6 Table 2f Network meta-analysis IOP estimates for drugs from studies published by 2014 ...... 32

7.2.7 Table 2g Network meta-analysis IOP estimates for classes from studies published by 1991 ...... 33

7.2.8 Table 2h Network meta-analysis IOP estimates for classes from studies published by 1995 ...... 34

7.2.9 Table 2i Network meta-analysis IOP estimates for classes from studies published by 2002 ...... 35

7.2.10 Table 2j Network meta-analysis IOP estimates for classes from studies published by 2004 ...... 36

7.2.11 Table 2k Network meta-analysis IOP estimates for classes from studies published by 2009 ...... 37

7.2.11 Table 2l Network meta-analysis IOP estimates for classes from studies published by 2014 ...... 38

7.3 Table 3. Guideline and network meta-analysis comparison ...... 39

8 Figures...... 40

8.1 Figure 1 Selection of studies ...... 40

8.2 Figure 2 Risk of bias figure ...... 41

vi

8.2.1 Figure 2a Risk of bias figure for studies published up to 1991 ...... 41

8.2.2 Figure 2b Risk of bias figure for studies published up to 1995 ...... 41

8.2.3 Figure 2c Risk of bias figure for studies published up to 2002 ...... 42

8.2.4 Figure 2d Risk of bias figure for studies published up to 2004 ...... 42

8.2.5 Figure 2e Risk of bias figure for studies published up to 2009 ...... 43

8.2.6 Figure 2f Risk of bias figure for studies published up to 2014 ...... 43

8.3 Figure 3 Network graphs ...... 44

8.3.1 Figure 3a Network graph for studies published up to 1991 ...... 44

8.3.2 Figure 3b Network graph for studies published up to 1995...... 44

8.3.3 Figure 3c Network graph for studies published up to 2002 ...... 45

8.3.4 Figure 3d Network graph for studies published up to 2004...... 45

8.3.5 Figure 3e Network graph for studies published up to 2009 ...... 46

8.3.6 Figure 3f Network graph for studies published up to 2014 ...... 46

8.4 Figure 4 Funnel plots of treatment effect relative to placebo at each network meta-analysis time point ...... 47

8.4.1 Figure 4a Funnel plot for drug effect relative to placebo ...... 47

8.4.2 Figure 4b Funnel plot for class effect relative to placebo ...... 48

8.5 Figure 5 Ranking probabilities for any treatment at any position ...... 49

8.5.1 Figure 5a Ranking probabilities for any drug at any position from studies published by 1991 ...... 49

8.5.2 Figure 5b Ranking probabilities for any drug at any position from studies published by 1995 ...... 50

8.5.3 Figure 5c Ranking probabilities for any drug at any position from studies published by 2002 ...... 51

8.5.4 Figure 5d Ranking probabilities for any drug at any position from studies published by 2004 ...... 52

8.5.5 Figure 5e Ranking probabilities for any drug at any position from studies published by 2009 ...... 53

vii

8.5.6 Figure 5f Ranking probabilities for any drug at any position from studies published by 2014 ...... 54

8.5.7 Figure 5g Ranking probabilities for any class at any position from studies published by 1991 ...... 55

8.5.8 Figure 5h Ranking probabilities for any class at any position from studies published by 1995 ...... 56

8.5.9 Figure 5i Ranking probabilities for any class at any position from studies published by 2002 ...... 57

8.5.10 Figure 5j Ranking probabilities for any class at any position from studies published by 2004 ...... 58

8.5.11 Figure 5k Ranking probabilities for any class at any position from studies published by 2009 ...... 59

8.5.12 Figure 5l Ranking probabilities for any class at any position from studies published by 2014 ...... 60

8.6 Figure 6 Cumulative ranking of treatments at each network meta-analysis time point… ...... 61

8.6.1 Figure 6a Cumulative ranking of drugs at each network meta-analysis time point………………… ...... 61

8.6.2 Figure 6b Cumulative ranking of class at each network meta-analysis time point…………...... 61

Appendix I. Search Strategy ...... 62

Appendix II. References to included studies ...... 68

Appendix III. Characteristics of included studies ...... 78

Appendix IV. Risk of bias table ...... 84

Appendix V. Pair-wise meta-analysis ...... 92 App V. Table 1. Summary estimates for intraocular pressure at 3 months derived from pair-wise meta-analysis for drugs in studies published by 1991 ...... 92 App V. Table 3. Summary estimates for intraocular pressure at 3 months derived from pair-wise meta-analysis for drugs in studies published by 2002 ...... 93

viii

App V. Table 2. Summary estimates for intraocular pressure at 3 months derived from pair-wise meta-analysis for drugs in studies published by 1995 ...... 94 App V. Table 4. Summary estimates for intraocular pressure at 3 months derived from pair-wise meta-analysis for drugs in studies published by 2004 ...... 95 App V. Table 5. Summary estimates for intraocular pressure at 3 months derived from pair-wise meta-analysis for drugs in studies published by 2009 ...... 96 App V. Table 6. Summary estimates for intraocular pressure at 3 months derived from pair-wise meta-analysis for drugs in studies published by 2014 ...... 97 App V. Table 7. Summary estimates for intraocular pressure at 3 months derived from pair-wise meta-analysis for classes in studies published by 1991 ...... 98 App V. Table 8. Summary estimates for intraocular pressure at 3 months derived from pair-wise meta-analysis for classes in studies published by 1995 ...... 99 App V. Table 8. Summary estimates for intraocular pressure at 3 months derived from pair-wise meta-analysis for classes in studies published by 2002 ...... 100 App V. Table 9. Summary estimates for intraocular pressure at 3 months derived from pair-wise meta-analysis for classes in studies published by 2004 ...... 101 App V. Table 10. Summary estimates for intraocular pressure at 3 months derived from pair-wise meta-analysis for classes in studies published by 2009 ...... 102 App V. Table 11. Summary estimates for intraocular pressure at 3 months derived from pair-wise meta-analysis for classes in studies published by 2014 ...... 103

Curriculum Vitae ...... 104

ix

1 Introduction

Clinical practice guidelines are statements of recommendations for patient care that are intended to be based on the best available evidence.1, 2 Historically, guidelines had primarily represented the opinions of individual authors or the consensus of experts.3

With the advent of evidence-based medicine, however, guidelines have increasingly made use of randomized controlled trials (RCTs) and synthesis of RCTs in the form of systematic reviews and meta-analyses to form the basis of recommendations.2 Despite the push towards evidence-based guidelines, there may still be many recommendations that are based on lower levels of evidence. Tricoci et al., for example, examined 17 recent cardiovascular guidelines and found that among the 16 guidelines that reported levels of evidence, recommendations were most frequently based on expert opinion, case studies, or standard of care.4 It is unknown in these cases whether no high quality evidence exists, evidence exists but authors were unaware of it, or advanced statistical methods were not available for them to address their questions.

When quantitatively evaluating the evidence base to make a guideline recommendation, the standard meta-analytic techniques may not always be adequate. A standard meta-analysis can only compare two treatments at a time, and only those treatments that have been compared directly in clinical trials. When developing a guideline for a particular condition, in many cases multiple treatment options must be considered, and direct comparisons may be available only for some pairs of treatments. In these cases, an alternative to the standard meta-analysis may be used, the network meta- analysis. A network meta-analysis looks across the entire network of trials of treatments for a specific condition and uses information from both direct and indirect comparisons

1

(i.e. using studies comparing treatments A and B and studies comparing B and C to estimate the comparison between A and C) to make inferences about the comparative effectiveness of all treatments in a single analysis.5,6 Since they enable an “all-way” comparison, network meta-analyses are particularly suited for informing evidence-based guideline recommendations.

Clinical fields which could most benefit from network meta-analysis are those for which a number of treatment options are available. One such area is primary open-angle glaucoma (POAG). POAG is an eye condition in which damage has occurred to the optic nerve and is associated with factors such as high intraocular pressure (IOP), age, and being African American.7 POAG makes up the majority of glaucoma cases.8 Since IOP is the only known modifiable risk factor for POAG, treatment efficacy is generally determined by reduction in IOP.7,9 One of the earliest sets of guidelines that has been influential in the care of POAG is the American Academy of Ophthalmology’s (AAO)

POAG Preferred Practice Pattern (PPP).7, 10-17 The first version of this guideline was published in 1989, with major revisions being published approximately every three to five years.

When the AAO PPP guideline was first developed, evidence was gathered based on the guideline panel members’ preexisting knowledge; each member submitted what they considered seminal works and these works were distributed among the rest of the panel.18 In 1996, the panel began using literature searching methods to gather evidence, though details of the search were not reported. The panel also began rating the strength of the evidence in three levels: “I” for evidence from RCTs, “II” for “an appropriately controlled case series and sufficient statistical analysis,” and “III” for “expert opinion.”12

2

In the 2000 publication, the panel started reporting more details about the literature search, such as databases searched and publication years included.13 The criteria for strength of the evidence was also revised. “I” represented “strong evidence in support of the statement” based on study design, study populations, general quality, and statistical methods.13 “II” represented “substantial evidence in support of the statement” based on lacking one or more of the components for level “I” categorization. The definition for

“III,” similar to before, represented a “consensus of expert opinion.”13 In 2010, the categorizations for strength of evidence were again redefined.17 “I” was for evidence based on high quality RCTs or meta-analyses. “II” represented evidence from well- designed non-randomized controlled trials, cohort studies, case-control studies, or multiple-time series studies. Support was rated as “III” for evidence from descriptive studies, case reports, or expert committee/organization reports.17

By using a cumulative network meta-analysis (i.e. conducting network meta- analysis on a collection of studies published up to a time point), the evidence base for first line medical treatments can be compared with the recommendations for treatment for each major revision in the AAO guideline. Findings from this study will inform guideline developers about the potential benefits of incorporating the results of network meta- analyses to form recommendations in the future. This study is not intended as criticism of guideline developers for not using statistical methods that were undeveloped at the time.

Rather, we would like to examine what impact such techniques would have had, had they been available at the time.

3

2 Objective

The objective of this study was to compare the clinical recommendations for first- line medical therapy for POAG from each major update of AAO’s POAG PPP with the actual evidence base as determined by network meta-analysis available at the time of each major update.

4

3 Methods

3.1 Guideline identification and extraction We identified nine versions of the AAO’s POAG PPP from the AAO website

(http://www.aao.org/preferred-practice-patterns-publication) from 1989 to 2010, updated about every three to five years. Since only the latest version could be obtained online, we contacted the AAO’s librarian who provided the remaining versions.19 One individual reviewed each version of the guideline, identified sections discussing treatment for

POAG, and extracted recommendations on specific drugs or drug classes for initial treatment, references cited for recommendations, and numerical estimates of efficacy or effectiveness (i.e. reduction of IOP) for drugs or drug classes. If a guideline included no specific recommendation for a drug or drug class, we extracted general recommendations for POAG management with medical therapies (e.g. “Medical therapy should be initial treatment for POAG”), as well as discussions about available medical therapies (e.g.

“Treatment A is most frequently prescribed as initial treatment”). We considered recommendations evidence-based if they were based on a least one high-quality large

RCT or a systematic review. When consecutive guideline versions presented identical recommendations or discussions regarding medical treatment, we grouped them together.

Therefore, the nine guidelines were divided into five groups based on their recommendations.

3.2 Systematic review and network meta-analysis This study was conducted using RCTs identified from an ongoing systematic review.20 We performed a network meta-analysis for each group of guidelines. Based on the latest guideline in each group, each corresponding network meta-analysis was comprised of all eligible studies published either up to the stopping point for the literature

5

search reported in the guideline or, if such a point is not reported, a year before that guideline was published (to allow for lag time between publication and inclusion of evidence in guideline). An additional analysis was performed with all studies obtained from the published literature up to 2014.

3.2.1 Eligibility criteria The eligibility criteria described here are the same as the underlying systematic review, unless otherwise noted.20 Eligible studies were RCTs with at least 60% of participants having a diagnosis of POAG or ocular hypertension (OHT), as defined by the trial. Trials included in this analysis also had to evaluate first line medical treatments for

POAG or OHT, and compared single active treatments with no treatment, placebo, or other single active treatments.

Trials were excluded if less than 10 participants were enrolled per treatment arm or if participants were followed for outcomes less than 28 days after randomization.

For this analysis, we examined mean IOP at 3 months as a continuous variable in units of mmHg as the primary outcome. When a trial measured IOP multiple ways, the priority for selection of IOP measurement was based on the following order: mean diurnal IOP, 24-hour mean IOP, peak IOP, morning IOP, and trough IOP. If a trial did not report IOP values at 3 months, we used data from the closest follow-up time point instead. IOP was selected as the primary outcome based on a preliminary analysis of guidelines indicating that it is the primary efficacy endpoint on which guideline recommendations were made.7

3.2.2 Identification of included studies We searched the Cochrane Register of Controlled Trials (CENTRAL) in The

Cochrane Library, MEDLINE, and EMBASE in November 17, 2009 and the search was

6

updated in March 11, 2014. Although the Food and Drug Administration (FDA) was searched for additional trials for the underlying systematic review, we did not include these trials in this project because none of the guidelines reported searching the FDA website. The search strategies are available in Appendix 1. Two individuals independently screened titles and abstracts of identified records for potential eligibility.

We obtained the full texts for records considered potentially eligible and these articles were then assessed independently for eligibility for the review by two individuals. When feasible, two individuals assessed the non-English language reports for eligibility, otherwise a single individual who was a native or fluent speaker of the language was responsible for assessing eligibility. We resolved discrepancies in classification of eligibility of full text articles through discussion or consultation with a third person.

3.2.3 Data abstraction and management Two individuals independently abstracted data from eligible trials on the study design, participant and intervention characteristics, outcomes, risk of bias, and quantitative results on treatment effects and safety using electronic forms developed and maintained in the Systematic Review Data Repository (http://srdr.ahrq.gov/).21,22 We used the Cochrane Risk of Bias Tool to grade each of the following methodological domains at “low” “high” or “unclear” risk of bias: sequence generation and allocation sequence concealment (selection bias), masking of participants and outcome assessors

(information bias), trial funding, and author financial relationships.23 We resolved discrepancies in data abstraction through discussion or consultation with a third person.

3.2.4 Qualitative synthesis We examined clinical, methodological, and statistical heterogeneity. We investigated clinical heterogeneity in terms of participant characteristics (e.g. age of

7

participants and baseline IOP) and trial interventions. For methodological heterogeneity, we considered study design and risk of bias.

3.2.5 Quantitative synthesis Our analysis did not distinguish between drug concentrations; comparisons were based on the active ingredient and class of that ingredient. We first conducted pair-wise meta-analyses for all direct comparisons using random-effects models assuming comparison-specific heterogeneity and a common heterogeneity across all comparisons at both the drug and class level. To assess the statistical heterogeneity, we examined the I2 and tau2 values for these models. Pair-wise meta-analyses were conducted in STATA

13®.24

Next, we fit Bayesian random-effects network meta-analysis models based on the

Lu and Ades approach in WinBUGS 1.4.3. 25-27 This model also accounts for the within- study correlation of multi-arm trials.25-26 We applied non-informative, yet proper, priors so that the data dominate the posterior distribution. We drew samples of the parameters of interest from the full posterior distribution using Markov Chain Monte Carlo (MCMC) algorithms. We used 2 chains and obtained 50,000 samples (after a 20,000 sample burn- in period). Our approach to model class effect was based on the approach used by Mayo-

Wilson et al.28 In this model, class effect is estimated from the pooled distribution of estimates from individual treatments in that class. This method allows us to use data from all trials for class effect, rather than discarding trials comparing drugs from the same class. We assumed that variance was homogeneous at both the drug and the class level.

3.2.6 Evaluation of network meta-analysis assumptions A valid network meta-analysis requires the assumption that there are no systematic differences between included comparisons other than the treatments

8

themselves.5 We examined this assumption based on the distribution of participant characteristics, interventions, and design characteristics among trials. We further considered the statistical disagreement between direct and indirect comparisons, or inconsistency, present among studies. To assess inconsistency, we used the loop-specific approach with inconsistency models. For the loop-specific approach, each independent closed triangular or quadratic loop (set of three or four treatments connected by direct comparisons) in the network is evaluated for inconsistency and incorporated as separate parameters (i.e. inconsistency factors) in the model.29-30 This analysis was conducted in

STATA 13®.30-32 When inconsistency was found, we qualitatively investigated trial characteristics such as funding source to determine potential sources of inconsistency.

3.2.7 Measures of relative treatment effect We examined mean differences in IOP (and 95% confidence intervals or credible intervals) between drug pairs and drug class pairs. We combined both change from baseline values with values at a time point. Due to randomization, it is reasonable to assume that both specific metrics are estimating the same effect.33 We also determined the probability of rank for each drug or class (i.e. the probability of a drug being the most efficacious treatment, the second most, etc.). We examined the hierarchy of treatment rankings by using the surface below the cumulative ranking curve (SUCRA).30,34 A

SUCRA value (or percentage) gives the probability that a treatment is among the best treatments, with a value of 1 (or 100%) meaning that a treatment is certain to be the best and 0 (or 0%) meaning that a treatment is certain to be the worst. Rankings based on

SUCRA values are considered to better take into account uncertainty in estimated treatment effects than general ranking probabilities.30,34

9

3.3 Guideline and network meta-analysis comparison We compared information extracted from each guideline group to the results of the corresponding network meta-analysis to assess frequency of matching of recommended drugs or drug classes and efficacy estimates in the guideline with the most efficacious drug or drug class from the network meta-analysis (based on SUCRA values).

10

4 Results

4.1 Guideline identification and extraction We identified 9 version of the AAO’s POAG PPP: 1989, 1990, 1992, 1996, 2000,

2003, 2005, 2006, and 2010.7,10-17 Based on recommendations and level of discussion of

POAG medical therapies, we grouped the guidelines together into 5 different sets: 1991-

1992, 1996, 2000-2003, 2005-2006, and 2010 (Table 1). Of these guideline sets, only

2010 made recommendations on first-line medical therapy. Based on a meta-analysis of

11 glaucoma trials, the 2010 guideline stated that “Prostaglandin analogs are the most effective drugs at lowering IOP and can be considered as initial medical therapy.”

However, no other guideline set made any specific recommendations with regard to which drug or class of drug is most efficacious; guideline statements have focused on describing available options, therapies most often used as first-line treatment, or general guidance for treatment. For example, the 2005-2006 guideline set stated that “In many instances, topical medications constitute effective initial therapy" instead of making a specific recommendation. Of the guideline sets, the 2005-2006 and 2010 sets reported stopping points for literature searches. Therefore, the time points for network meta- analysis were 1991, 1995, 2002, 2004, and 2009, with an additional one comprising all collected data up to 2014.

4.2 Network meta-analysis 4.2.1 Search results and general study characteristics We identified 10,936 unique records from the search. For this analysis, a total of

105 RCTs from the published literature met our eligibility criteria (Figure 1; references for these trials are available in Appendix II). The first trial was published in 1983 and the latest trial in 2013. The network included 18 trials (1,161 participants) by 1991, 29 trials

11

(2,641 participants) by 1995, 66 trials (9,446 participants) by 2002, 76 trials (10,717 participants) by 2004, and 91 trials (13,870 participants) by 2009. As of 2014, there are a total 16,898 participants in the network of 105 trials. Detailed characteristics of individual trials are included in Appendix III.

The study characteristics described include all trials published by the network meta-analysis time point. Sample size of studies appears to be smaller in earlier years than later ones. In 1991, the median size of trials was 69 participants (interquartile range

(IQR): 28 to 85). This increased to 72 participants (IQR: 42 to 137) by 1995, and 95 participants (IQR: 45 to 177) by 2002. Afterwards, study sample size does not appear to change substantially, with a median of 91 participants (IQR: 43 to 195) by 2004 and 90 participants (IQR: 47 to 213) by 2009. As of 2014, the median sample size of trials is 97

(IQR: 49 to 218). The smallest trial (17 participants) was published in 1985 and the largest (976 participants) in 2005.

The proportion of trials reported to be multicenter also appears to be smaller in earlier years: 39% of trials were reported to be multicenter in 1991, 55% in 1995, 70% in

2002, 64% in 2004, 61% in 2009, and 65% as of 2014. Reported regions of participant recruitment (in 1991 and 2014 respectively) are North America (28% to 37%), Latin

America (0% to 3%), Europe (0% to 17%), Africa (0% to 1%), Asia (6% to 16%),

Oceania (0% to 2%) (trials could recruit participants from more than one region; remaining trials did not report region).

The length of trials is generally longer for earlier studies. Median trial length was

6 months (IQR: 3 to 15) in 1991 and 6 months (IQR: 3 to 12) in 1995. For all network meta-analysis time points after 1995, median length was 3 months (IQR: 3 to 12).

12

4.2.2 Risk of bias At all network meta-analysis time points, the risk of bias of included studies was generally unclear to high, although the proportion of trials with low risk of bias appears to be higher at later time points (Figure 2a-f; Appendix IV). The proportion of studies with low risk of bias from sequence generation ranges from 11% in 1991 (89% unclear) to 43% in 2014 (57% unclear). 17% of studies published by 1991 were rated to have a low risk of bias for allocation concealment (83% unclear) while 27% published by 2014 were (73% unclear). We rated the risk of bias from masking of participants to be low for

33% of studies up to 1991 (11 % high; 56% unclear) and 39% of studies up to 2014 (21% high; 40% unclear). 17% of studies to 1991 (83% unclear) were rated to have a low risk of bias due to masking of IOP assessor, and 22% of studies were rated low up to 2014

(10% high; 69% unclear). In trials published up to 1991, only 33% reported funding, of which 100% had industry funding and 33% were funded by government (a trial could report more than one funding source). By 2014, 59% reported funding, of which 92% reported industry funding and 13% reported government funding. Of trials published by

1991, 33% reported on author financial conflicts of interest, of which 100% reported conflicts of interest for at least one author. By 2014, 52% of trials reported on conflicts of interests, of which 67% reported existing conflict of interest for least one author.

4.2.3 Interventions Included trials studied 13 active interventions from 4 different classes, as well as placebo/vehicle/no treatment (Figure 3a-f). The active interventions were apraclonidine and brimonidine (alpha-2 adrenergic agonists); betaxolol, carteolol, levobunolol, and timolol (beta blockers); brinzolamide and dorzolamide (carbonic anhydrase inhibitors);

13

and bimatoprost, latanoprost, travoprost, tafluprost, and unoprostone (prostaglandin analogs). By 1991, three active drugs (betaxolol, levobunolol, and timolol) from one class

(beta blockers) and placebo were studied in RCTs. In 1995, the network of studies included an additional five drugs (apraclonidine, carteolol, dorzolamide, latanoprost, and unoprostone), and at least one drug from each class was included. The network expanded to 11 active drugs in 2002 with brimonidine, brinzolamide, and travoprost. By 2004, 12 of the active drugs were in the network and no additional drugs were added in 2009. As of 2014, one more drug, tafluprost, has been studied in the network.

4.2.4 Network meta-analysis outcomes Network meta-analysis indicates that all drugs (and classes) are superior to placebo in lowering 3-month IOP at all network meta-analysis time points (Table 2a-l;

Figure 4a-b). Results are reported in terms of mean IOP (in mmHg) and 95% credible interval. The drugs and classes with the largest effect on IOP reduction compared with placebo at each time point are: 1991: levobunolol 4.53 (3.31 to 5.79), beta blockers 4.01

(0.48 to 7.43); 1995: apraclonidine 5.63 (2.56 to 8.64), alpha agonists 5.64 (1.73 to 9.50);

2002: travoprost 6.02 (4.64 to 7.38), prostaglandins 4.97 (3.29 to 6.65); 2004: bimatoprost 5.87 (4.67 to 7.06), prostaglandins 4.75 (3.11 to 6.44); 2009 bimatoprost

5.87 (4.96 to 6.77), prostaglandins 4.58 (2.94 to 6.24); 2014: bimatoprost 5.55 (4.80 to

6.31), prostaglandins 4.38 (3.03 to 5.75). Point estimates for drug and class effects appear to diminish over time (Figure 4a-b).

Many direct comparisons between drugs, such as latanoprost vs placebo, are missing even by 2014 and for the direct comparisons that exist, there are often only one or two trials (Appendix V). The class effect estimates from direct comparison differ greatly from those obtained from combining direct and indirect comparisons. For

14

example, by 2014 only two trials have directly compared prostaglandins with placebo and the pooled estimate is not significantly superior to placebo.

Ranking probabilities are consistent with the network meta-analysis effect estimates (Figure 5a-l). By 2014, for example, bimatoprost had a 93.4% chance of being the most efficacious drug in terms of effect on 3-month IOP, 6.1% chance of being the second best, and 0.5% chance of being the third best, while as a class, prostaglandins had

73.9% chance of being the best, 19.8% of being the second best, and 5.4% of being the third best. Ranking based on cumulative ranking probabilities from SUCRA plots are also generally consistent with effect estimates (Figure 6a-b). The only time at which the highest cumulative rank did not match with treatment effect was in 1995, in which apraclonidine was had the highest mean effect but levobunolol had the highest cumulative ranking. By 2004, rankings generally stabilized for both drugs and classes.

Sometimes, when two drugs were included at the same time point, they crossed in cumulative rank at subsequent points (Figure 6a-b). For example, from 2002 to 2009, brimonidine was ranked higher than timolol, but by 2014, their positions switched.

4.2.5 Inconsistency By 2014, the loop-specific approach to inconsistency indicated evidence of inconsistency in 5 of 34 triangular loops (15%). We could not find any qualitative reasons to explain inconsistency among studies included in the inconsistent loops.

4.3 Guideline and network meta-analysis comparison A summary of the comparison between guidelines recommendations and network meta-analytic findings is given in Table 3. Based on network-meta-analysis, it would have been possible to make treatment recommendations for all guideline sets, such as that beta blockers were superior to placebo/no treatment based on RCT data available by

15

1991. The only time evidence from network meta-analysis with the guideline recommendation is in 2010, as both indicate that the prostaglandin class should be considered the first-line treatment in terms of efficacy.

16

5 Discussion

AAO’s POAG PPP did not make specific recommendations for first-line treatment until 2010. However, based on network meta-analysis, there was sufficient evidence to conclude that medical treatments superior to placebo existed by 1991. The

2010 recommendation is supported by the network meta-analysis results. The ranking of classes based on the effect sizes given in the 2010 guidelines are also consistent with our findings. Prostaglandins have the largest effect, beta blockers and alpha agonists are next and are very close in effect, and carbonic anhydrase inhibitors are the least effective. In both the 2010 guideline and the corresponding network meta-analaysis, even though prostaglandins are considered the most efficacious class, the magnitude of IOP reduction does not really appear to differ substantially between classes.

The AAO’s POAG PPPs do not give recommendations at the drug level. This may be because the guideline producers did not want to appear to favor a particular drug manufacturer, since some glaucoma drugs, such as bimatoprost, are still under patent.

Our results indicate that drugs within a class generally have similar effects on IOP. The most notable exception is unoprostone, which was the least effective drug at all time points since 2004 despite the high ranking of all other prostaglandins. With unoprostone, there is uncertainty whether it should be classified as a prostaglandin analogue or not.35-36

Despite being derived from prostaglandin F(2α) like the other prostaglandin drugs, pharmacological studies have suggested that unoprostone has a distinct mechanism of action compared to the other prostaglandins, and therefore it may not be appropriate to group it with these other drugs.35-36 If unoprostone is really part of a separate drug class, it would explain the great disparity in IOP effect, as well as indicate that our findings for

17

prostaglandin effect may underestimate the true class effect. Other exceptions are betaxolol, which has a lower effect than the other beta blockers, and the two alpha agonists, apraclonidine and brimonidine, which start off with different effectiveness profiles but appear to get closer in effect size and ranking over time. Since within-class treatments are generally similar, it is appropriate for guidelines to make recommendations at the class level for POAG treatments.

One interesting finding from the cumulative network meta-analysis is that, as can be seen in Figure 4, there appears to be a consistent pattern that the effect size for all glaucoma treatments diminishes over time. Despite this, all treatments are superior to placebo at all network meta-analysis time points. This result is consistent with findings by Gehr et al., who, in a meta-regression, determined that the effect size for both timolol and latanoprost decreased over time.37 One potential explanation for this finding is due to small-study effects; the tendency for smaller studies to produce larger treatment effects than larger studies due to factors like publication bias or poorer methodological quality of smaller studies.38 Earlier studies in the network were smaller, but even when median study size stops increasing in 2002, treatment effect size still diminishes. Another possibility is that in earlier studies, participants had less severe disease or more easily controlled IOP. After drugs became established in RCTs, people entering trials may be those with more severe disease or whose IOP was not controlled on initial therapy. In

Gehr et al.’s study, it was found that a significant relationship existed between baseline

IOP values and treatment effect over time in the case of timolol.37 We will further explore the data to try and determine the decrease in effect size over time.

18

The AAO PPPs used IOP as the major determinant in forming treatment recommendations and so did our analysis, yet it is widely understood that IOP is a surrogate outcome for visual function.39 Visual field is considered an outcome more clinically relevant to visual function, but requires longer follow-up time than IOP to accurately assess changes (generally years).39 Based on our included studies, median follow-up time in POAG trials is 3 months by 2014, preventing meaningful assessment of visual field. Some studies, such as the recent UK Glaucoma Treatment Study, which was conducted to assess whether latanoprost preserves visual field in addition to reducing

IOP, have indicated that IOP and visual field are associated.40 On the other hand, trials such as the Low-Pressure Glaucoma Treatment Study, which found that participants assigned to brimonidine had better preserved visual field than those on timolol despite mean IOP being similar in both groups, have suggested that depending on IOP is questionable.39,41 An additional concern is that even if IOP is demonstrated to be a reliable predictor for visual field for treatments in one class, different classes may affect visual field progression differently despite having a similar ocular hypotensive effect.39 In terms of the guidelines, the AAO POAG PPPs have considered the evidence associating

IOP with risk of visual field progression to be sufficient that IOP is an acceptable outcome for trials since 1996.7,13-17

Cumulative pair-wise meta-analysis has demonstrated the importance of using meta-analysis instead of just looking at individual RCTs to inform treatment recommendations. A cumulative meta-analysis by Antman et al. showed that sufficient

RCT evidence existed to confirm that thrombolytic therapy significantly reduced the risk of death from myocardial infarction by 1973, but that it took 13 years (by which the

19

number of RCTs had increased from 10 to 43) for the therapy to be recommended for routine practice by expert reviewers.42

Network meta-analysis has begun to be recognized as a useful tool for guideline developers. The Endocrine Society commissioned a network meta-analysis to be conducted to inform recommendations for its 2012 clinical practice guideline for osteoporosis in men.43-44 The National Institute for Health and Clinical Excellence

(NICE) in Europe also conducted its own network meta-analysis for making recommendations on neuropathic pain treatments.45

By extending the principles of cumulative meta-analysis to network meta- analysis, we have provided evidence that network meta-analysis can benefit clinical guideline developers. If network meta-analysis results had been available to developers, the POAG PPP could have made recommendations for initial medical treatment at each major update. Furthermore, the current first-line treatment, prostaglandins, could have been recommended as early as the 2003 update, seven years earlier than when prostaglandins were recommended. Another strength of our analysis is that we were able to estimate class effect without discarding trials comparing drugs from the same class.

Our findings regarding network meta-analysis and guideline comparisons may not be applicable to other clinical fields or even other glaucoma guidelines such as those developed by the National Institute for Health and Clinical Excellence,9 since we only examined a single set of guidelines. The cultures of other clinical fields or different glaucoma guideline groups may lead them to have different approaches to making treatment recommendations (e.g. drug level instead of class level) or in their use of evidence as the basis for recommendations.

20

Conclusion

We identified 5 sets of guidelines from AAO’s POAG PPPs with major revisions in terms of medical treatment recommendations or discussions of therapies. Treatment recommendations were only made in the final set. Using cumulative network meta- analysis, we were able to determine the best drug and class at the time of each major revision based on RCT evidence available at the time. Both the final guideline and the corresponding network meta-analysis indicate that prostaglandins should be considered first-line treatment in terms of IOP reduction. Other findings from the network meta- analysis are that the effect size for all drugs and classes appears to decrease over time, but all were significantly better than placebo at all time points. Network meta-analysis results have the potential to help clinical guideline developers make evidence-based recommendations.

21

6 References

1. Institute of Medicine. Clinical Practice Guidelines We Can Trust. 2011 2. Institute of Medicine. Knowing What Works in Healthcare: Roadmap for the Nation. 2008 3. Eddy DM. Evidence-based medicine: a unified approach. Health Aff (Millwood). 2005;24(1):9–17. 4. Tricoci P, Allen JM, Kramer JM, Califf RM, Smith SC. Scientific evidence underlying the ACC/AHA clinical practice guidelines. JAMA. 2009;301(8):831– 41. 5. Salanti G. Indirect and mixed-treatment comparison, network, or multiple- treatments meta-analysis: many names, many benefits, many concerns for the next generation evidence synthesis tool. Res Synth Methods. 2012;3(2):80–97. 6. Cipriani. 2013. Conceptual and technical challenges in network meta-analysis. 7. American Academy of Ophthalmology. Primary Open-Angle Glaucoma Preferred Practice Patterns. San Francisco: American Academy of Ophthalmology; 2010 8. Quigley HA, Broman AT. The number of people with glaucoma worldwide in 2010 and 2020. Br J Ophthalmol. 2006;90:262-7. 9. National Institute for Health and Care Excellence. NICE guidelines. Glaucoma: diagnosis and management of chronic open angle glaucoma and ocular hypertension. 2009. 10. American Academy of Ophthalmology. Primary Open-Angle Glaucoma Preferred Practice Patterns. San Francisco: American Academy of Ophthalmology; 1989 11. American Academy of Ophthalmology. Primary Open-Angle Glaucoma Preferred Practice Patterns. San Francisco: American Academy of Ophthalmology; 1990 12. American Academy of Ophthalmology. Primary Open-Angle Glaucoma Preferred Practice Patterns. San Francisco: American Academy of Ophthalmology; 1992 13. American Academy of Ophthalmology. Primary Open-Angle Glaucoma Preferred Practice Patterns. San Francisco: American Academy of Ophthalmology; 1996

22

14. American Academy of Ophthalmology. Primary Open-Angle Glaucoma Preferred Practice Patterns. San Francisco: American Academy of Ophthalmology; 2000 15. American Academy of Ophthalmology. Primary Open-Angle Glaucoma Preferred Practice Patterns. San Francisco: American Academy of Ophthalmology; 2003 16. American Academy of Ophthalmology. Primary Open-Angle Glaucoma Preferred Practice Patterns. San Francisco: American Academy of Ophthalmology; 2005 17. American Academy of Ophthalmology. Primary Open-Angle Glaucoma Preferred Practice Patterns. San Francisco: American Academy of Ophthalmology; 2006 18. Sommer A, Weiner JP, Gamble L. Developing specialty-wide standards of practice: the experience of ophthalmology. QRB Qual Rev Bull. 1990;16(2):65– 70. 19. "The American Academy of Ophthalmology." American Academy of Ophthalmology. Accessed 21 Aug. 2014. http://www.aao.org/ 20. Li T, Lindsey K, Rouse B, et al. Comparative Effectiveness of First-line Medications for Patients with Primary Open Angle Glaucoma or Ocular Hypertension – A Systematic Review and Network Meta-analysis (in process) 21. Ip S, Hadar N, Keefe S, Parkin C, Iovin R, Balk EM, Lau J. A Web-based archive of systematic review data. Syst Rev. 2012;1(1):15. 22. Li T, Vedula SS, Hadar N, Parkin C, Lau J, Dickersin K. Innovations in data collection, management, and archiving for systematic reviews. Ann Intern Med. 2015;162(4):287-94. 23. Higgins JPT, Altman DG, Sterne JAC (editors). Chapter 8: Assessing risk of bias in included studies. In: Higgins JPT, Green S (editors). Cochrane Handbook for Systematic Reviews of Interventions Version 5.1.0 (updated March 2011). The Cochrane Collaboration, 2011. Available from www.cochrane-handbook.org 24. StataCorp. 2013. Stata Statistical Software: Release 13. College Station, TX: StataCorp LP. 25. Lu G, Ades AE. Combination of direct and indirect evidence in mixed treatment comparisons. Stat Med. 2004;23(20):3105-3124. 26. Lu G, Ades a. E. Assessing Evidence Inconsistency in Mixed Treatment Comparisons. J Am Stat Assoc. 2006;101(474):447–459.

23

27. Lunn DJ, Thomas A, Best N, Spiegelhalter D. WinBUGS— a Bayesian modelling framework: concepts, structure, and extensibility. Stat Comput. 2000; 10: 325–37. 28. Mayo-Wilson E, Dias S, Mavranezouli I, et al. Psychological and pharmacological interventions for social anxiety disorder in adults: a systematic review and network meta-analysis. The Lancet Psychiatry. 2014;1(5):368–376. 29. Dias S, Welton NJ, Sutton AJ, Caldwell DM, Lu G, Ades AE. Evidence synthesis for decision making 4: inconsistency in networks of evidence based on randomized controlled trials. Med Decis Making. 2013;33(5):641-656. 30. Chaimani A, Higgins JPT, Mavridis D, Spyridonos P, Salanti G. Graphical tools for network meta-analysis in STATA. PLoS One. 2013;8(10):e76654. 31. White IR. Multivariate random-effects meta-regression : Updates to mvmeta. Stata J. 2011;11(2):255-270. 32. White IR, Barrett JK, Jackson D, Higgins JPT. Consistency and inconsistency in network meta-analysis: model estimation using multivariate meta-regression. Res Synth Methods. 2012;3(2):111–125. 33. Deeks JJ, Higgins JPT, and Altman DG on behalf of the Cochrane Statistical Methods Group. Chapter 9.4.5.2 Meta-analysis of change scores. In: Higgins JPT, Green S (editors). Cochrane Handbook for Systematic Reviews of Interventions Version 5.1.0 (updated March 2011). The Cochrane Collaboration, 2011. Available from www.cochrane-handbook.org; accessed on April 29, 2015 34. Salanti G, Ades AE, Ioannidis JP. Graphical methods and numerical summaries for presenting results from multiple-treatment meta-analysis: an overview and tutorial. J Clin Epidemiol. 2011;64(2):163–71. 35. Harms N V, Toris CB. Current status of unoprostone for the management of glaucoma and the future of its use in the treatment of retinal disease. Expert Opin Pharmacother. 2013;14(1):105-113. 36. Fung DS, T WJ. An evidence-based review of unoprostone for glaucoma : place in therapy. Clin Ophthalmol. 2014;8:543-554. 37. Gehr BT, Weiss C, Porzsolt F. The fading of reported effectiveness. A meta- analysis of randomised controlled trials. BMC Med Res Methodol. 2006;6:25. 38. Sterne J AC, Gavaghan D, Egger M. Publication and related bias in meta- analysis: Power of statistical tests and prevalence in the literature. J Clin Epidemiol. 2000;53(11):1119-1129.

24

39. Medeiros FA. Biomarkers and surrogate endpoints in glaucoma clinical trials. Br J Ophthalmol. 2014:1–5. 40. Garway-Heath DF, Crabb DP, Bunce C, et al. Latanoprost for open-angle glaucoma (UKGTS): a randomised, multicentre, placebo-controlled trial. Lancet. 2014:1295-1304. 41. Krupin T, Liebmann JM, Greenfield DS, Ritch R, Gardiner S. A randomized trial of brimonidine versus timolol in preserving visual function: Results from the low-pressure glaucoma treatment study. Am J Ophthalmol. 2011;151(4):671-681. 42. Antman EM, Lau J, Kupelnick B, Mosteller F, Chalmers TC. A comparison of results of meta-analyses of randomized control trials and recommendations of clinical experts. Treatments for myocardial infarction. JAMA. 1992;268(2):240- 248 43. Watts NB, Adler R a, Bilezikian JP, et al. Osteoporosis in men: an Endocrine Society clinical practice guideline. J Clin Endocrinol Metab. 2012;97(6):1802- 1822. 44. Murad MH, Drake MT, Mullan RJ, et al. Clinical review. Comparative effectiveness of drug treatments to prevent fragility fractures: a systematic review and network meta-analysis. J Clin Endocrinol Metab. 2012;97(6):1871- 1880. 45. National Institute for Health and Care Excellence. NICE guidelines. Neuropathic pain – pharmacological management: The pharmacological management of neuropathic pain in adults in non-specialist settings. 2013.

25

7 Tables

7.1 Table 1 Recommendations from AAO POAG PPPs

26

7.2 Table 2 Summary estimates for intraocular pressure (mmHg) at 3 months derived from network meta-analysis 7.2.1 Table 2a Network meta-analysis IOP estimates for drugs from studies published by 1991

27

7.2.2 Table 2b Network meta-analysis IOP estimates for drugs from studies published by 1995

28

7.2.3 Table 2c Network meta-analysis IOP estimates for drugs from studies published by 2002

29

7.2.4 Table 2d Network meta-analysis IOP estimates for drugs from studies published by 2004

30

7.2.5 Table 2e Network meta-analysis IOP estimates for drugs from studies published by 2009

31

7.2.6 Table 2f Network meta-analysis IOP estimates for drugs from studies published by 2014

32

7.2.7 Table 2g Network meta-analysis IOP estimates for classes from studies published by 1991

33

7.2.8 Table 2h Network meta-analysis IOP estimates for classes from studies published by 1995

34

7.2.9 Table 2i Network meta-analysis IOP estimates for classes from studies published by 2002

35

7.2.10 Table 2j Network meta-analysis IOP estimates for classes from studies published by 2004

36

7.2.11 Table 2k Network meta-analysis IOP estimates for classes from studies published by 2009

37

7.2.11 Table 2l Network meta-analysis IOP estimates for classes from studies published by 2014

Color coding: drug class Mean difference < 0 favors the drug in the column Mean difference > 0 favors the drug in the row Reported numbers are calculated by column - row under the Lu and Ades homogeneous random effects model assuming consistency Reported posterior means and 95% Bayesian credible intervals

Grey Placebo/vehicle/no treatment Orange Alpha-2 adrenergic agonist Green Beta-blocker Red Carbonic anhydrase inhibitor Blue Prostaglandin analog

38

7.3 Table 3. Guideline and network meta-analysis comparison

39

8 Figures

8.1 Figure 1 Selection of studies

40

8.2 Figure 2 Risk of bias figure 8.2.1 Figure 2a Risk of bias figure for studies published up to 1991

8.2.2 Figure 2b Risk of bias figure for studies published up to 1995

41

8.2.3 Figure 2c Risk of bias figure for studies published up to 2002

8.2.4 Figure 2d Risk of bias figure for studies published up to 2004

42

8.2.5 Figure 2e Risk of bias figure for studies published up to 2009

8.2.6 Figure 2f Risk of bias figure for studies published up to 2014

43

8.3 Figure 3 Network graphs 8.3.1 Figure 3a Network graph for studies published up to 1991

Drugs Classes

Betaxolol

Beta blockers

Levobunolol Placebo

Placebo Timolol # of studies = 18 # of patients contributing to this network = 1161

8.3.2 Figure 3b Network graph for studies published up to 1995

Drugs Classes

Betaxolol Alpha agonists Carteolol Beta blockers Apraclonidine

Levobunolol

Placebo Placebo

Timolol

Unoprostone Carbonic anhydrase inhibitors

Dorzolamide Latanoprost Prostaglandins # of studies = 29 (1 three-arm study) # of patients contributing to this network = 2641

44

8.3.3 Figure 3c Network graph for studies published up to 2002

Drugs Classes

Betaxolol Alpha agonists Carteolol Brimonidine Beta blockers

Levobunolol Apraclonidine

Timolol Placebo Placebo

Brinzolamide Unoprostone Carbonic anhydrase inhibitors

Dorzolamide Travoprost Latanoprost # of studies = 66 (5 three-arm studies) Prostaglandins # of patients contributing to this network = 9446

8.3.4 Figure 3d Network graph for studies published up to 2004

Drugs Classes Betaxolol Carteolol Brimonidine Alpha agonists Levobunolol Beta blockers

Apraclonidine Timolol

Placebo Placebo

Brinzolamide

Unoprostone Carbonic anhydrase inhibitors Dorzolamide Travoprost Bimatoprost Prostaglandins Latanoprost # of studies = 76 (7 three-arm studies) # of patients contributing to this network = 10717

45

8.3.5 Figure 3e Network graph for studies published up to 2009

Drugs Classes

Betaxolol Carteolol Alpha agonists Brimonidine Beta blockers Levobunolol

Apraclonidine Timolol

Placebo Placebo

Brinzolamide

Unoprostone

Dorzolamide Carbonic anhydrase inhibitors Travoprost Bimatoprost Latanoprost # of studies = 91 (11 three-arm studies) Prostaglandins # of patients contributing to this network = 13870

8.3.6 Figure 3f Network graph for studies published up to 2014

Drugs Classes

Carteolol Betaxolol Alpha agonists Levobunolol Brimonidine Beta blockers Timolol Apraclonidine

Brinzolamide Placebo Placebo

Dorzolamide Unoprostone

Carbonic anhydrase inhibitors Bimatoprost Tafluprost

Latanoprost Travoprost # of studies = 105 (12 three-arm studies) Prostaglandins # of patients contributing to this network = 19562 Each node represents one drug. The drugs are color-coded by class. The size of the node is proportional to the number of participants randomized to that drug.

The edges represent direct comparisons (i.e. when there is a line connecting two drugs, the two drugs have been compared directly to each other in a trial). The width of the edge is proportional to the number of trials.

Grey Placebo/vehicle/no treatment Orange Alpha-2 adrenergic agonist Green Beta-blocker Red Carbonic anhydrase inhibitor Blue Prostaglandin analog

46

8.4 Figure 4 Funnel plots of treatment effect relative to placebo at each network meta-analysis time point 8.4.1 Figure 4a Funnel plot for drug effect relative to placebo

47

8.4.2 Figure 4b Funnel plot for class effect relative to placebo

Since glaucoma drugs are expected to lower IOP values, more negative IOP values indicate greater effect.

48

8.5 Figure 5 Ranking probabilities for any treatment at any position 8.5.1 Figure 5a Ranking probabilities for any drug at any position from studies published by 1991

49

8.5.2 Figure 5b Ranking probabilities for any drug at any position from studies published by 1995

50

8.5.3 Figure 5c Ranking probabilities for any drug at any position from studies published by 2002

51

8.5.4 Figure 5d Ranking probabilities for any drug at any position from studies published by 2004

52

8.5.5 Figure 5e Ranking probabilities for any drug at any position from studies published by 2009

53

8.5.6 Figure 5f Ranking probabilities for any drug at any position from studies published by 2014

54

8.5.7 Figure 5g Ranking probabilities for any class at any position from studies published by 1991

100% 80% 60% 0.982 40% Rank 2 20% Rank 1 0% 0.018

55

8.5.8 Figure 5h Ranking probabilities for any class at any position from studies published by 1995

100% 90% 80% 70% 60% 50% 40% Rank 5 30% 0.517 20% 0.318 Rank 4 10% Rank 3 0.099 0.066 0% 0.000 Rank 2 Rank 1

56

8.5.9 Figure 5i Ranking probabilities for any class at any position from studies published by 2002

100% 90% 80% 70% 60% 50% 40% 0.680 30% Rank 5 20% Rank 4 Rank 3 10% 0.170 0.124 Rank 2 0% 0.000 0.027 Rank 1

57

8.5.10 Figure 5j Ranking probabilities for any class at any position from studies published by 2004

100% 90% 80% 70% 60% 50%

40% 0.708 30% Rank 5 20% Rank 4 10% 0.144 0.105 Rank 3 0% 0.000 0.043 Rank 2 Rank 1

58

8.5.11 Figure 5k Ranking probabilities for any class at any position from studies published by 2009

59

8.5.12 Figure 5l Ranking probabilities for any class at any position from studies published by 2014

100% 90% 80% 70% 60%

50% Rank 5 40% Rank 4 0.739 Rank 3 30% Rank 2 20% Rank 1 10% 0.111 0.119 0% 0.000

Warmer colors indicate better ranks

60

8.6 Figure 6 Cumulative ranking of treatments at each network meta- analysis time point 8.6.1 Figure 6a Cumulative ranking of drugs at each network meta-analysis time point

8.6.2 Figure 6b Cumulative ranking of class at each network meta-analysis time point

SUCRA percentage is the probability a treatment has of being among the best treatments (e.g. 100% if certainly the best, 0% if certainly the worst)

61

Appendix I. Search Strategy

Cochrane Library #1 MeSH descriptor: [Glaucoma, Open-Angle] explode all trees #2 MeSH descriptor: [Ocular Hypertension] explode all trees #3 (open near/2 angle near/2 glaucoma*) #4 (POAG or OHT) #5 (((increas* or elevat* or high*) near/3 (ocular or intra-ocular)) and pressure) #6 {or #1-#5} #7 MeSH descriptor: [Adrenergic beta-Antagonists] explode all trees #8 MeSH descriptor: [Timolol] explode all trees #9 Timolol* #10 MeSH descriptor: [Metipranolol] explode all trees #11 Metipranolol* #12 MeSH descriptor: [Carteolol] explode all trees #13 Carteolol* #14 MeSH descriptor: [Levobunolol] explode all trees #15 Levobunolol* #16 MeSH descriptor: [Betaxolol] explode all trees #17 Betaxolol* #18 MeSH descriptor: [Carbonic Anhydrase Inhibitors] explode all trees #19 (Carbonic near/2 Anhydrase near/2 Inhibitor*) #20 MeSH descriptor: [Acetazolamide] explode all trees #21 Acetazolam* #22 Brinzolamide* #23 Dorzolamide* #24 MeSH descriptor: [Prostaglandins, Synthetic] explode all trees #25 latanoprost* #26 travoprost* #27 bimatoprost* #28 unoprostone* #29 tafluprost* #30 MeSH descriptor: [Antihypertensive Agents] explode all trees #31 MeSH descriptor: [Pilocarpine] explode all trees #32 Pilocarpin* #33 MeSH descriptor: [Epinephrine] explode all trees #34 epinephrine* #35 dipivefrin* #36 MeSH descriptor: [Adrenergic alpha-2 Receptor Agonists] explode all trees 62

#37 (adrenergic near/2 alpha* near/3 agonist*) #38 apraclonidin* #39 brimonidine* #40 (drug* or medic* or pharmacologic*) near/3 (treat* or therap* or intervent*) #41 {or #7-#40} #42 #6 and #41

MEDLINE (OVID) 1. exp clinical trial/ [publication type] 2. (randomized or randomised).ab,ti. 3. placebo.ab,ti. 4. dt.fs. 5. randomly.ab,ti. 6. trial.ab,ti. 7. groups.ab,ti. 8. or/1-7 9. exp animals/ 10. exp humans/ 11. 9 not (9 and 10) 12. 8 not 11 13. exp glaucoma open angle/ 14. exp ocular hypertension/ 15. (open adj2 angle adj2 glaucoma$).tw. 16. (POAG or OHT).tw. 17. (((increas$ or elevat$ or high$) adj3 (ocular or intra-ocular)) and pressure).tw.

18. or/13-17 19. exp adrenergic beta antagonists/ 20. exp timolol/ 21. timolol$.tw. 22. exp metipranolol/ 23. metipranolol$.tw. 24. exp carteolol/ 25. carteolol$.tw. 26. exp levobunolol/ 27. levobunolol$.tw. 28. exp betaxolol/ 29. betaxolol$.tw.

63

30. exp carbonic anhydrase inhibitors/ 31. (carbonic adj2 anhydrase adj2 inhibitor$).tw. 32. exp Acetazolamide/ 33. acetazolamide$.tw. 34. brinzolamide$.tw. 35. dorzolamide$.tw. 36. exp Prostaglandins, Synthetic/ 37. latanoprost$.tw. 38. travoprost$.tw. 39. bimatoprost$.tw. 40. unoprostone$.tw. 41. brimonidine$.tw. 42. exp antihypertensive agents/ 43. exp pilocarpine/ 44. pilocarpin$.tw. 45. exp epinephrine/ 46. epinephrin$.tw. 47. dipivefrin$.tw. 48. exp Adrenergic alpha-2 Receptor Agonists/ 49. ((adrenergic adj2 alpha$ adj2 receptor$) or (adrenergic adj2 alpha$ adj2 agonist$)).tw. 50. apraclonidin$.tw. 51. tafluprost$.tw. 52. ((drug$ or medic$ or pharmacologic$) adj3 (treat$ or therap$ or intervent$)).tw. 53. or/19-52 54. 18 and 53 55. 12 and 54

Embase.com #1 'randomized controlled trial'/exp #2 'randomization'/exp #3 'double blind procedure'/exp #4 'single blind procedure'/exp #5 random*:ab,ti #6 #1 OR #2 OR #3 OR #4 OR #5 #7 'animal'/exp OR 'animal experiment'/exp #8 'human'/exp #9 #7 AND #8 #10 #7 NOT #9

64

#11 #6 NOT #10 #12 'clinical trial'/exp #13 (clin* NEAR/3 trial*):ab,ti #14 ((singl* OR doubl* OR trebl* OR tripl*) NEAR/3 (blind* OR mask*)):ab,ti #15 'placebo'/exp #16 placebo*:ab,ti #17 random*:ab,ti #18 'experimental design'/exp #19 'crossover procedure'/exp #20 'control group'/exp #21 'latin square design'/exp #22 #12 OR #13 OR #14 OR #15 OR #16 OR #17 OR #18 OR #19 OR #20 OR #21 #23 #22 NOT #10 #24 #23 NOT #11 #25 'comparative study'/exp #26 'evaluation'/exp #27 'prospective study'/exp #28 control*:ab,ti OR prospectiv*:ab,ti OR volunteer*:ab,ti #29 #25 OR #26 OR #27 OR #28 #30 #29 NOT #10 #31 #30 NOT (#11 OR #23) #32 #11 OR #24 OR #31 #33 'open angle glaucoma'/exp #34 'intraocular hypertension'/exp #35 (open NEAR/2 angle):ab,ti AND (angle NEAR/2 glaucoma*):ab,ti #36 poag:ab,ti OR oht:ab,ti #37 ((increas* OR elevat* OR high*) NEAR/3 (ocular OR 'intra ocular')):ab,ti AND pressure:ab,ti #38 #33 OR #34 OR #35 OR #36 OR #37 #39 'beta adrenergic receptor blocking agent'/exp #40 'timolol'/exp #41 timolol*:ab,ti #42 'metipranolol'/exp #43 metipranolol*:ab,ti #44 'carteolol'/exp #45 carteolol*:ab,ti #46 'levobunolol'/exp #47 levobunolol*:ab,ti

65

#48 'betaxolol'/exp #49 betaxolol*:ab,ti #50 'carbonate dehydratase inhibitor'/exp #51 (carbonic NEAR/2 anhydrase):ab,ti AND (anhydrase NEAR/2 inhibitor*):ab,ti #52 'acetazolamide'/exp #53 acetazolamide*:ab,ti #54 brinzolamide*:ab,ti #55 dorzolamide*:ab,ti #56 'latanoprost'/exp #57 latanoprost*:ab,ti #58 'travoprost'/exp #59 travoprost*:ab,ti #60 'bimatoprost'/exp #61 bimatoprost*:ab,ti #62 'unoprostone isopropyl ester'/exp #63 unoprostone*:ab,ti #64 'brimonidine'/exp #65 brimonidine*:ab,ti #66 'antihypertensive agent'/exp #67 'pilocarpine'/exp #68 pilocarpin*:ab,ti #69 'adrenalin'/exp #70 epinephrin*:ab,ti #71 dipivefrin*:ab,ti #72 'alpha 2 adrenergic receptor stimulating agent'/exp #73 (adrenergic NEAR/2 alpha*):ab,ti AND (alpha* NEAR/2 agonist*):ab,ti #74 apraclonidin*:ab,ti #75 'tafluprost'/exp #76 tafluprost*:ab,ti #77 ((drug* OR medic* OR pharmacologic*) NEAR/3 (treat* OR therap* OR intervent*)):ab,ti #78 #39 OR #40 OR #41 OR #42 OR #43 OR #44 OR #45 OR #46 OR #47 OR #48 OR #49 OR #50 OR #51 OR #52 OR #53 OR #54 OR #55 OR #56 OR #57 OR #58 OR #59 OR #60 OR #61 OR #62 OR #63 OR #64 OR #65 OR #66 OR #67 OR #68 OR #69 OR #70 OR #71 OR #72 OR #73 OR #74 OR #75 OR #76 OR #77 #79 #38 AND #78 #80 #32 AND #79

66

PubMed #1 ((randomized controlled trial[pt]) OR (controlled clinical trial[pt]) OR (randomised[tiab] OR randomized[tiab]) OR (placebo[tiab]) OR (drug therapy[sh]) OR (randomly[tiab]) OR (trial[tiab]) OR (groups[tiab])) NOT (animals[mh] NOT humans[mh]) #2 (open[tw] AND angle[tw] AND glaucoma*[tw]) NOT Medline[sb] #3 (POAG[tw] OR OHT[tw]) NOT Medline[sb] #4 (((increase*[tw] OR elevat*[tw] OR high*[tw]) AND (ocular[tw] OR intra-ocular[tw])) AND pressure[tw]) NOT Medline[sb] #5 #2 OR #3 OR #4 #6 timolol*[tw] NOT Medline[sb] #7 metipranolol*[tw] NOT Medline[sb] #8 carteolol*[tw] NOT Medline[sb] #9 levobunolol*[tw] NOT Medline[sb] #10 betaxolol*[tw] NOT Medline[sb] #11 (carbonic[tw] AND anhydrase[tw] AND inhibitor*[tw]) NOT Medline[sb] #12 acetazolamide*[tw] NOT Medline[sb] #13 brinzolamide*[tw] NOT Medline[sb] #14 dorzolamide*[tw] NOT Medline[sb] #15 latanoprost*[tw] NOT Medline[sb] #16 travoprost*[tw] NOT Medline[sb] #17 bimatoprost*[tw] NOT Medline[sb] #18 unoprostone*[tw] NOT Medline[sb] #19 brimonidine*[tw] NOT Medline[sb] #20 pilocarpin*[tw] NOT Medline[sb] #21 epinephrin*[tw] NOT Medline[sb] #22 dipivefrin* NOT Medline[sb] #23 ((adrenergic[tw] AND alpha*[tw] AND receptor*[tw]) OR (adrenergic[tw] AND alpha*[tw] AND agonist*[tw])) NOT Medline[sb] #24 apraclonidin*[tw] NOT Medline[sb] #25 tafluprost*[tw] NOT Medline[sb] #26 ((drug*[tw] OR medic*[tw] OR pharmacologic*[tw]) AND (treat*[tw] OR therap*[tw] OR intervent*[tw])) NOT Medline[sb] #27 #6 OR #7 OR #8 OR #9 OR #10 OR #11 OR #12 OR #13 OR #14 OR #15 OR #16 OR #17 OR #18 OR #19 OR #20 OR #21 OR #22 OR #23 OR #24 OR #25 OR #26 #28 #5 AND #27 #29 #1 AND #2

67

Appendix II. References to included studies

1. Radius RL. Use of betaxolol in the reduction of elevated intraocular pressure. Arch Ophthalmol 1983; 101(6): 898-900. 2. Berry DP, Jr., Van Buskirk EM, Shields MB. Betaxolol and timolol. A comparison of efficacy and side effects. Arch Ophthalmol 1984; 102(1): 42-5. 3. Bensinger RE, Keates EU, Gofman JD, Novack GD, Duzman E. Levobunolol. A three-month efficacy study in the treatment of glaucoma and ocular hypertension. Arch Ophthalmol 1985; 103(3): 375-8. 4. Berson FG, Cohen HB, Foerster RJ, Lass JH, Novack GD, Duzman E. Levobunolol compared with timolol for the long-term control of elevated intraocular pressure. Arch Ophthalmol 1985; 103(3): 379-82. 5. Cinotti A, Cinotti D, Grant W, et al. Levobunolol vs timolol for open- angle glaucoma and ocular hypertension. Am J Ophthalmol 1985; 99(1): 11-7. 6. Ober M, Scharrer A, David R, et al. Long-term ocular hypotensive effect of levobunolol: results of a one-year study. Br J Ophthalmol 1985; 69(8): 593-9. 7. Stryz JR, Merte HJ. [Pressure lowering effect and side effects of 0.5% and 1.0% levobunolol eyedrops, compared with 0.5% timolol eyedrops in patients with open-angle glaucoma]. Klin Monbl Augenheilkd 1985; 187(6): 537-44. 8. Stewart RH, Kimbrough RL, Ward RL. Betaxolol vs timolol. A six- month double-blind comparison. Arch Ophthalmol 1986; 104(1): 46-8. 9. Boozman FW, 3rd, Carriker R, Foerster R, Allen RC, Novack GD, Batoosingh AL. Long-term evaluation of 0.25% levobunolol and timolol for therapy for elevated intraocular pressure. Arch Ophthalmol 1988; 106(5): 614-8. 10. Feghali JG, Kaufman PL, Radius RL, Mandell AI. A comparison of betaxolol and timolol in open angle glaucoma and ocular hypertension. Acta Ophthalmol (Copenh) 1988; 66(2): 180-6. 11. Freyler H, Novack GD, Menapace R, Skorpik C, Mordaunt J, Batoosingh AL. [Comparison of the effectiveness and safety of levobunolol and timolol in ocular hypertension and chronic open-angle glaucoma]. Klin Monbl Augenheilkd 1988; 193(3): 257-60. 12. Long DA, Johns GE, Mullen RS, et al. Levobunolol and betaxolol. A double-masked controlled comparison of efficacy and safety in patients with elevated intraocular pressure. Ophthalmology 1988; 95(6): 735-41.

68

13. Seamone C, LeBlanc R, Saheb N, Novack G. Efficacy of twice-daily levobunolol in the treatment of elevated intraocular pressure. Can J Ophthalmol 1988; 23(4): 168-70. 14. Epstein DL, Krug JH, Jr., Hertzmark E, Remis LL, Edelstein DJ. A long-term clinical trial of timolol therapy versus no treatment in the management of glaucoma suspects. Ophthalmology 1989; 96(10): 1460- 7. 15. Kass MA, Gordon MO, Hoff MR, et al. Topical timolol administration reduces the incidence of glaucomatous damage in ocular hypertensive individuals. A randomized, double-masked, long-term clinical trial. Arch Ophthalmol 1989; 107(11): 1590-8. 16. Yoshiaki K, Ikuo A, Makoto A. Clinical evaluation of betaxolol hydrochloride in the treatment of primary open angle glaucoma and ocular hypertension. Multi-center double-masked study in comparison with timolol. Rinsho Hyoka (Clinical Evaluation) 1989; 17(2): 243-74. 17. Schulzer M, Drance SM, Douglas GR. A comparison of treated and untreated glaucoma suspects. Ophthalmology 1991; 98(3): 301-7. 18. Silverstone D, Zimmerman T, Choplin N, et al. Evaluation of once-daily levobunolol 0.25% and timolol 0.25% therapy for increased intraocular pressure. Am J Ophthalmol 1991; 112(1): 56-60. 19. Beehler CC, Stewart WC, Macdonald DK, et al. A comparison of the ocular hypotensive efficacy of twice-daily 0.25% levobunolol to 0.5% timolol in patients previously treated with 0.5% timolol. J Glaucoma 1992; 1(4): 237-42. 20. Flammer J, Kitazawa Y, Bonomi L, et al. Influence of carteolol and timolol on IOP an visual fields in glaucoma: a multi-center, double- masked, prospective study. Eur J Ophthalmol 1992; 2(4): 169-74. 21. Azuma I, Masuda K, Kitazawa Y, Takase M, Yamamura H. Double- masked comparative study of UF-021 and timolol ophthalmic solutions in patients with primary open-angle glaucoma or ocular hypertension. Jpn J Ophthalmol 1993; 37(4): 514-25. 22. Nagasubramanian S, Hitchings RA, Demailly P, et al. Comparison of apraclonidine and timolol in chronic open-angle glaucoma. A three- month study. Ophthalmology 1993; 100(9): 1318-23. 23. Wilkerson M, Cyrlin M, Lippa EA, et al. Four-week safety and efficacy study of dorzolamide, a novel, active topical carbonic anhydrase inhibitor. Arch Ophthalmol 1993; 111(10): 1343-50. 24. Behrens-Baumann W, Kimmich F, Walt JG, Lue J. A comparison of the ocular hypotensive efficacy and systemic safety of 0.5% levobunolol and 2% carteolol. Ophthalmologica 1994; 208(1): 32-6.

69

25. Kitazawa Y. Phase III comparative study of MK-507 ophthalmic solution in primary open-angle glaucoma and ocular hypertension. Folia Ophthalmol Jpn 1994; 45(9): 1023-33. 26. Ravalico G, Salvetat L, Toffoli G, Pastori G, Croce M, Battaglia P. Ocular hypertension: A follow-up study in treated and untreated patients. NEW TRENDS OPHTHALMOL 1994; 9(2): 97-101. 27. Alm A, Stjernschantz J. Effects on intraocular pressure and side effects of 0.005% latanoprost applied once daily, evening or morning. A comparison with timolol. Scandinavian Latanoprost Study Group. Ophthalmology 1995; 102(12): 1743-52. 28. Schwartz B, Lavin P, Takamoto T, Araujo DF, Smits G. Decrease of optic disc cupping and pallor of ocular hypertensives with timolol therapy. Acta Ophthalmol Scand Suppl 1995; (215): 5-21. 29. Strahlman E, Tipping R, Vogel R. A double-masked, randomized 1-year study comparing dorzolamide (Trusopt), timolol, and betaxolol. International Dorzolamide Study Group. Arch Ophthalmol 1995; 113(8): 1009-16. 30. Fristrom B. A 6-month, randomized, double-masked comparison of latanoprost with timolol in patients with open angle glaucoma or ocular hypertension. Acta Ophthalmol Scand 1996; 74(2): 140-4. 31. Mishima HK, Masuda K, Kitazawa Y, Azuma I, Araie M. A comparison of latanoprost and timolol in primary open-angle glaucoma and ocular hypertension. A 12-week study. Arch Ophthalmol 1996; 114(8): 929-32. 32. Schuman JS. Clinical experience with brimonidine 0.2% and timolol 0.5% in glaucoma and ocular hypertension. Surv Ophthalmol 1996; 41 Suppl 1: S27-37. 33. Serle JB. A comparison of the safety and efficacy of twice daily brimonidine 0.2% versus betaxolol 0.25% in subjects with elevated intraocular pressure. The Brimonidine Study Group III. Surv Ophthalmol 1996; 41 Suppl 1: S39-47. 34. Stewart WC, Laibovitz R, Horwitz B, Stewart RH, Ritch R, Kottler M. A 90-day study of the efficacy and side effects of 0.25% and 0.5% apraclonidine vs 0.5% timolol. Apraclonidine Primary Therapy Study Group. Arch Ophthalmol 1996; 114(8): 938-42. 35. Watson P, Stjernschantz J. A six-month, randomized, double-masked study comparing latanoprost with timolol in open-angle glaucoma and ocular hypertension. The Latanoprost Study Group. Ophthalmology 1996; 103(1): 126-37. 36. Yamamoto T, Kitazawa Y, Noma A, et al. The effects of the beta- adrenergic-blocking agents, timolol and carteolol, on plasma lipids and lipoproteins in Japanese glaucoma patients. J Glaucoma 1996; 5(4): 252-7.

70

37. Kitazawa Y, Azuma I, Shirato S, et al. Phase III Clinical Study of AG- 901 Ophthalmic Solution on Primary Open-Angle Glaucoma and Ocular Hypertension: A Multicenter, Double-Blind Comparison with 0.5% Timolol Maleate. Journal of Clinical Therapeutics and Medicines 1997; 13(11): 2975-91. 38. Stewart WC, Cohen JS, Netland PA, Weiss H, Nussbaum LL. Efficacy of carteolol hydrochloride 1% vs timolol maleate 0.5% in patients with increased intraocular pressure. Nocturnal Investigation of Glaucoma Hemodynamics Trial Study Group. Am J Ophthalmol 1997; 124(4): 498- 505. 39. Boyle JE, Ghosh K, Gieser DK, Adamsons IA. A randomized trial comparing the dorzolamide-timolol combination given twice daily to monotherapy with timolol and dorzolamide. Dorzolamide-Timolol Study Group. Ophthalmology 1998; 105(10): 1945-51. 40. Clineschmidt CM, Williams RD, Snyder E, Adamsons IA. A randomized trial in patients inadequately controlled with timolol alone comparing the dorzolamide-timolol combination to monotherapy with timolol or dorzolamide. Dorzolamide-Timolol Combination Study Group. Ophthalmology 1998; 105(10): 1952-9. 41. Diestelhorst M, Almegard B. Comparison of two fixed combinations of latanoprost and timolol in open-angle glaucoma. Graefes Arch Clin Exp Ophthalmol 1998; 236(8): 577-81. 42. LeBlanc RP. Twelve-month results of an ongoing randomized trial comparing brimonidine tartrate 0.2% and timolol 0.5% given twice daily in patients with glaucoma or ocular hypertension. Brimonidine Study Group 2. Ophthalmology 1998; 105(10): 1960-7. 43. Rusk C, Sharpe E, Laurence J, Polis A, Adamsons I. Comparison of the efficacy and safety of 2% dorzolamide and 0.5% betaxolol in the treatment of elevated intraocular pressure. Dorzolamide Comparison Study Group. Clin Ther 1998; 20(3): 454-66. 44. Silver LH. Clinical efficacy and safety of brinzolamide (Azopt), a new topical carbonic anhydrase inhibitor for primary open-angle glaucoma and ocular hypertension. Brinzolamide Primary Therapy Study Group. Am J Ophthalmol 1998; 126(3): 400-8. 45. Bojic L, Bagatin J, Ivanisevic M, Hozo I, Racic G, Karelovic D. Influence of betaxolol and timolol on the venous tone in glaucoma patients. Int Ophthalmol 1999; 23(3): 149-53. 46. Stewart WC, Dubiner HB, Mundorf TK, et al. Effects of carteolol and timolol on plasma lipid profiles in older women with ocular hypertension or primary open-angle glaucoma. Am J Ophthalmol 1999; 127(2): 142-7.

71

47. Toris CB, Camras CB, Yablonski ME. Acute versus chronic effects of brimonidine on aqueous humor dynamics in ocular hypertensive patients. Am J Ophthalmol 1999; 128(1): 8-14. 48. Alm A, Widengard I. Latanoprost: experience of 2-year treatment in Scandinavia. Acta Ophthalmol Scand 2000; 78(1): 71-6. 49. O'Donoghue EP. A comparison of latanoprost and dorzolamide in patients with glaucoma and ocular hypertension: a 3 month, randomised study. Ireland Latanoprost Study Group. Br J Ophthalmol 2000; 84(6): 579-82. 50. Sall K. The efficacy and safety of brinzolamide 1% ophthalmic suspension (Azopt) as a primary therapy in patients with open-angle glaucoma or ocular hypertension. Brinzolamide Primary Therapy Study Group. Surv Ophthalmol 2000; 44 Suppl 2: S155-62. 51. Bron AM, Denis P, Nordmann JP, Rouland JF, Sellem E, Johansson M. Additive IOP-reducing effect of latanoprost in patients insufficiently controlled on timolol. Acta ophthalmologica Scandinavica 2001; 79(3): 289-93. 52. DuBiner HB, Mroz M, Shapiro AM, Dirks MS, Brimonidine vs. Latanoprost Study G. A comparison of the efficacy and tolerability of brimonidine and latanoprost in adults with open-angle glaucoma or ocular hypertension: a three-month, multicenter, randomized, double- masked, parallel-group trial. Clin Ther 2001; 23(12): 1969-83. 53. Kobayashi H, Kobayashi K, Okinami S. A comparison of intraocular pressure-lowering effect of prostaglandin F2 -alpha analogues, latanoprost, and unoprostone isopropyl. J Glaucoma 2001; 10(6): 487- 92. 54. Susanna R, Jr., Giampani J, Jr., Borges AS, Vessani RM, Jordao ML. A double-masked, randomized clinical trial comparing latanoprost with unoprostone in patients with open-angle glaucoma or ocular hypertension. Ophthalmology 2001; 108(2): 259-63. 55. Aung T, Chew PT, Oen FT, et al. Additive effect of unoprostone and latanoprost in patients with elevated intraocular pressure. Br J Ophthalmol 2002; 86(1): 75-9. 56. Bergstrand IC, Heijl A, Harris A. Dorzolamide and ocular blood flow in previously untreated glaucoma patients: a controlled double-masked study. Acta Ophthalmol Scand 2002; 80(2): 176-82. 57. Halpern MT, Covert DW, Robin AL. Projected impact of travoprost versus both timolol and latanoprost on visual field deficit progression and costs among black glaucoma subjects. Trans Am Ophthalmol Soc. 2002;100:109-17 58. Fellman RL, Sullivan EK, Ratliff M, et al. Comparison of travoprost 0.0015% and 0.004% with timolol 0.5% in patients with elevated

72

intraocular pressure: a 6-month, masked, multicenter trial. Ophthalmology 2002; 109(5): 998-1008. 59. Higginbotham EJ, Feldman R, Stiles M, Dubiner H, Fixed Combination Investigative G. Latanoprost and timolol combination therapy vs monotherapy: one-year randomized trial. Arch Ophthalmol 2002; 120(7): 915-22. 60. Jampel HD, Bacharach J, Sheu WP, et al. Randomized clinical trial of latanoprost and unoprostone in patients with elevated intraocular pressure. Am J Ophthalmol 2002; 134(6): 863-71. 61. Kampik A, Arias-Puente A, O'Brart DP, Vuori ML, European Latanoprost Study G. Intraocular pressure-lowering effects of latanoprost and brimonidine therapy in patients with open-angle glaucoma or ocular hypertension: a randomized observer-masked multicenter study. J Glaucoma 2002; 11(2): 90-6. 62. Nordmann JP, Mertz B, Yannoulis NC, et al. A double-masked randomized comparison of the efficacy and safety of unoprostone with timolol and betaxolol in patients with primary open-angle glaucoma including pseudoexfoliation glaucoma or ocular hypertension. 6 month data. Am J Ophthalmol 2002; 133(1): 1-10. 63. Pfeiffer N, European Latanoprost Fixed Combination Study G. A comparison of the fixed combination of latanoprost and timolol with its individual components. Graefes Arch Clin Exp Ophthalmol 2002; 240(11): 893-9. 64. Simmons ST, Earl ML, Alphagan/Xalatan Study G. Three-month comparison of brimonidine and latanoprost as adjunctive therapy in glaucoma and ocular hypertension patients uncontrolled on beta- blockers: tolerance and peak intraocular pressure lowering. Ophthalmology 2002; 109(2): 307-14; discussion 14-5. 65. Sponsel WE, Paris G, Trigo Y, Pena M. Comparative effects of latanoprost (Xalatan) and unoprostone (Rescula) in patients with open- angle glaucoma and suspected glaucoma. Am J Ophthalmol 2002; 134(4): 552-9. 66. Tsukamoto H, Mishima HK, Kitazawa Y, et al. A comparative clinical study of latanoprost and isopropyl unoprostone in Japanese patients with primary open-angle glaucoma and ocular hypertension. J Glaucoma 2002; 11(6): 497-501. 67. Camras CB, Hedman K, Group USLS. Rate of response to latanoprost or timolol in patients with ocular hypertension or glaucoma. J Glaucoma 2003; 12(6): 466-9. 68. Cardascia N, Vetrugno M, Trabucco T, Cantatore F, Sborgia C. Effects of travoprost eye drops on intraocular pressure and pulsatile ocular blood flow: a 180-day, randomized, double-masked comparison with

73

latanoprost eye drops in patients with open-angle glaucoma. Current therapeutic research, clinical and experimental 2003; 64(7): 389-400. 69. Inan UU, Ermis SS, Yucel A, Ozturk F. The effects of latanoprost and brimonidine on blood flow velocity of the retrobulbar vessels: a 3-month clinical trial. Acta Ophthalmol Scand 2003; 81(2): 155-60. 70. Kamal D, Garway-Heath D, Ruben S, et al. Results of the betaxolol versus placebo treatment trial in ocular hypertension. Graefes Arch Clin Exp Ophthalmol 2003; 241(3): 196-203. 71. Parrish RK, Palmberg P, Sheu WP, Group XLTS. A comparison of latanoprost, bimatoprost, and travoprost in patients with elevated intraocular pressure: a 12-week, randomized, masked-evaluator multicenter study. Am J Ophthalmol 2003; 135(5): 688-703. 72. Erkin EF, Tarhan S, Kayikcioglu OR, Deveci H, Guler C, Goktan C. Effects of betaxolol and latanoprost on ocular blood flow and visual fields in patients with primary open-angle glaucoma. Eur J Ophthalmol 2004; 14(3): 211-9. 73. Kobayashi H, Iwakiri R, Kobayashi K, Okinami S. Hypotensive effect of unoprostone as adjunct to latanoprost during multiple drug therapy for glaucoma. Japanese Journal of Clinical Ophthalmology 2004; 58(2): 193-7. 74. Vetrugno M, Cardascia N, Cantatore F, Sborgia C. Comparison of the effects of bimatoprost and timolol on intraocular pressure and pulsatile ocular blood flow in patients with primary open-angle glaucoma: A prospective, open-label, randomized, two-arm, parallel-group study. Current therapeutic research, clinical and experimental 2004; 65(6): 444-54. 75. Walters TR, DuBiner HB, Carpenter SP, Khan B, VanDenburgh AM, Bimatoprost Circadian IOPSG. 24-Hour IOP control with once-daily bimatoprost, timolol gel-forming solution, or latanoprost: a 1-month, randomized, comparative clinical trial. Surv Ophthalmol 2004; 49 Suppl 1: S26-35. 76. Wang TH, Huang JY, Hung PT, Shieh JW, Chen YF. Ocular hypotensive effect and safety of brinzolamide ophthalmic solution in open angle glaucoma patients. J Formos Med Assoc 2004; 103(5): 369- 73. 77. Barnebey HS, Orengo-Nania S, Flowers BE, et al. The safety and efficacy of travoprost 0.004%/timolol 0.5% fixed combination ophthalmic solution. Am J Ophthalmol 2005; 140(1): 1-7. 78. Camras CB, Sheu WP, United States Latanoprost-Brimonidine Study G. Latanoprost or brimonidine as treatment for elevated intraocular pressure: multicenter trial in the United States. J Glaucoma 2005; 14(2): 161-7.

74

79. Miglior S, Zeyen T, Pfeiffer N, et al. Results of the European Glaucoma Prevention Study. Ophthalmology 2005; 112(3): 366-75. 80. Erkin EF, Celik P, Kayikcioglu O, Deveci HM, Sakar A. Effects of latanoprost and betaxolol on cardiovascular and respiratory status of newly diagnosed glaucoma patients. Ophthalmologica 2006; 220(5): 332-7. 81. Koz OG, Ozsoy A, Yarangumeli A, Kose SK, Kural G. Comparison of the effects of travoprost, latanoprost and bimatoprost on ocular circulation: a 6-month clinical trial. Acta Ophthalmol Scand 2007; 85(8): 838-43. 82. Martin E, Martinez-de-la-Casa JM, Garcia-Feijoo J, Troyano J, Larrosa JM, Garcia-Sanchez J. A 6-month assessment of bimatoprost 0.03% vs timolol maleate 0.5%: hypotensive efficacy, macular thickness and flare in ocular-hypertensive and glaucoma patients. Eye (Lond) 2007; 21(2): 164-8. 83. Alagoz G, Gurel K, Bayer A, Serin D, Celebi S, Kukner S. A comparative study of bimatoprost and travoprost: effect on intraocular pressure and ocular circulation in newly diagnosed glaucoma patients. Ophthalmologica 2008; 222(2): 88-95. 84. Brandt JD, Cantor LB, Katz LJ, et al. Bimatoprost/timolol fixed combination: a 3-month double-masked, randomized parallel comparison to its individual components in patients with glaucoma or ocular hypertension. J Glaucoma 2008; 17(3): 211-6. 85. Kaback M, Scoper SV, Arzeno G, et al. Intraocular pressure-lowering efficacy of brinzolamide 1%/timolol 0.5% fixed combination compared with brinzolamide 1% and timolol 0.5%. Ophthalmology 2008; 115(10): 1728-34, 34 e1-2. 86. Williams RD, Cohen JS, Gross RL, et al. Long-term efficacy and safety of bimatoprost for intraocular pressure lowering in glaucoma and ocular hypertension: year 4. Br J Ophthalmol 2008; 92(10): 1387-92. 87. Yildirim N, Sahin A, Gultekin S. The effect of latanoprost, bimatoprost, and travoprost on circadian variation of intraocular pressure in patients with open-angle glaucoma. J Glaucoma 2008; 17(1): 36-9. 88. Casson RJ, Liu L, Graham SL, et al. Efficacy and safety of bimatoprost as replacement for latanoprost in patients with glaucoma or ocular hypertension: a uniocular switch study. Journal of glaucoma 2009; 18(8): 582-8. 89. Prata TS, Piassi MV, Melo LA, Jr. Changes in visual function after intraocular pressure reduction using antiglaucoma medications. Eye (Lond) 2009; 23(5): 1081-5. 90. Sharma R, Kohli K, Kapoor B, Mengi RK, Sadhotra P. The cardio- vascular effects of topical timolol, levobunolol and betaxolol in patients

75

of chronic simple glaucoma. Journal of Clinical and Diagnostic Research 2009; 3(4): 1615-20. 91. Zhao WJ. [Comparison of travoprost and timolol in patients with primary open angle glaucoma and ocular hypertension]. International Journal of Ophthalmology 2009; 9(9): 1753-4. 92. Birt CM, Buys YM, Ahmed, II, Trope GE. Prostaglandin efficacy and safety study undertaken by race (the PRESSURE study). Journal of glaucoma 2010; 19(7): 460-7. 93. Craven ER, Liu CC, Batoosingh A, Schiffman RM, Whitcup SM. A randomized, controlled comparison of macroscopic conjunctival hyperemia in patients treated with bimatoprost 0.01% or vehicle who were previously controlled on latanoprost. Clin Ophthalmol 2010; 4: 1433-40. 94. Higginbotham EJ, Olander KW, Kim EE, Grunden JW, Kwok KK, Tressler CS. Fixed combination of latanoprost and timolol vs individual components for primary open-angle glaucoma or ocular hypertension: a randomized, double-masked study. Archives of ophthalmology 2010; 128(2): 165-72. 95. Kammer JA, Katzman B, Ackerman SL, Hollander DA. Efficacy and tolerability of bimatoprost versus travoprost in patients previously on latanoprost: a 3-month, randomised, masked-evaluator, multicentre study. The British journal of ophthalmology 2010; 94(1): 74-9. 96. Macky TA. Bimatoprost versus travoprost in an Egyptian population: a hospital-based prospective, randomized study. Journal of ocular pharmacology and therapeutics : the official journal of the Association for Ocular Pharmacology and Therapeutics 2010; 26(6): 605-10. 97. Traverso CE, Ropo A, Papadia M, Uusitalo H. A phase II study on the duration and stability of the intraocular pressure-lowering effect and tolerability of Tafluprost compared with latanoprost. Journal of ocular pharmacology and therapeutics : the official journal of the Association for Ocular Pharmacology and Therapeutics 2010; 26(1): 97-104. 98. Nie X, Yang XH, Qin XF, Quan CJ, Yang SS. Correlation between the development of high myopia and intraocular pressure. International Journal of Ophthalmology 2011; 11(6): 1092-4. 99. Zhao L, Wang YL, Meng ZY, Hong H. Efficacy and safety of domestic Latanoprost in treating with open angle glaucoma and ocular hypertension. [Chinese]. International Journal of Ophthalmology 2011; 11(11): 1973-5. 100. Araie M, Yamazaki Y, Sugiyama K, Kuwayama Y, Tanihara H. [Phase III clinical trial of brimonidine in patients with primary open-angle glaucoma and ocular hypertension--comparison of the effects of brimonidine monotherapy versus timolol monotherapy, or combination

76

brimonidine/prostaglandins therapy versus combination placebo/prostaglandins therapy]. Nippon Ganka Gakkai Zasshi - Acta Societatis Ophthalmologicae Japonicae 2012; 116(10): 955-66. 101. Chabi A, Varma R, Tsai JC, et al. Randomized clinical trial of the efficacy and safety of preservative-free tafluprost and timolol in patients with open-angle glaucoma or ocular hypertension. American journal of ophthalmology 2012; 153(6): 1187-96. 102. Crichton AC, Vold S, Williams JM, Hollander DA. Ocular surface tolerability of prostaglandin analogs and prostamides in patients with glaucoma or ocular hypertension. Advances in Therapy 2013; 30(3): 260-70. 103. Delval L, Baudouin C, Gabisson P, Alliot E, Vincent B. Safety and efficacy of unpreserved timolol 0.1% gel in patients controlled by preserved latanoprost with signs of ocular intolerance. Journal français d'ophtalmologie 2013; 36(4): 316-23. 104. Katz G, Dubiner H, Samples J, Vold S, Sall K. Three-month randomized trial of fixed-combination brinzolamide, 1%, and brimonidine, 0.2%. JAMA Ophthalmology 2013; 131(6): 724-30. 105. Nguyen QH, McMenemy MG, Realini T, Whitson JT, Goode SM. Phase 3 randomized 3-month trial with an ongoing 3-month safety extension of fixed-combination brinzolamide 1%/brimonidine 0.2%. Journal of Ocular Pharmacology and Therapeutics 2013; 29(3): 290-7.

77

Appendix III. Characteristics of included studies

78

79

80

81

82

83

Appendix IV. Risk of bias table

84

85

86

87

88

89

90

91

Appendix V. Pair-wise meta-analysis

App V. Table 1. Summary estimates for intraocular pressure at 3 months derived from pair-wise meta-analysis for drugs in studies published by 1991

92

App V. Table 3. Summary estimates for intraocular pressure at 3 months derived from pair-wise meta-analysis for drugs in studies published by 2002

93

App V. Table 2. Summary estimates for intraocular pressure at 3 months derived from pair-wise meta-analysis for drugs in studies published by 1995

94

App V. Table 4. Summary estimates for intraocular pressure at 3 months derived from pair-wise meta-analysis for drugs in studies published by 2004

95

App V. Table 5. Summary estimates for intraocular pressure at 3 months derived from pair-wise meta-analysis for drugs in studies published by 2009

96

App V. Table 6. Summary estimates for intraocular pressure at 3 months derived from pair-wise meta-analysis for drugs in studies published by 2014

97

App V. Table 7. Summary estimates for intraocular pressure at 3 months derived from pair-wise meta-analysis for classes in studies published by 1991

98

App V. Table 8. Summary estimates for intraocular pressure at 3 months derived from pair-wise meta-analysis for classes in studies published by 1995

99

App V. Table 8. Summary estimates for intraocular pressure at 3 months derived from pair-wise meta-analysis for classes in studies published by 2002

100

App V. Table 9. Summary estimates for intraocular pressure at 3 months derived from pair-wise meta-analysis for classes in studies published by 2004

101

App V. Table 10. Summary estimates for intraocular pressure at 3 months derived from pair-wise meta-analysis for classes in studies published by 2009

102

App V. Table 11. Summary estimates for intraocular pressure at 3 months derived from pair-wise meta-analysis for classes in studies published by 2014

a Mean difference is calculated using the intraocular pressure of the treatment in column 2 - column 1 Mean difference > 0 favors the drug in column 1 Mean difference < 0 favors the drug in column 2 Color coding: Grey Placebo/vehicle/no treatment Orange Alpha-2 adrenergic agonist Green Beta-blocker Red Carbonic anhydrase inhibitor Blue Prostaglandin analog

103

Curriculum Vitae Benjamin Rouse 7926 Starburst Drive, Pikesville, MD 21208

(410) 653-3412 [email protected]

EDUCATION Master of Health Sciences (MHS) in Epidemiology Expected May 2015 Johns Hopkins Bloomberg School of Public Health, Baltimore, MD Concentration: Clinical Trials and Evidence Synthesis G.P.A.: 3.93/4.00

University of Illinois at Chicago College of Pharmacy, Chicago, IL Graduate-level coursework in Medicinal Chemistry and Pharmacognosy G.P.A.: 4.00/4.00

Bachelor of Liberal Arts & Science May 2014 Harriet L. Wilkes Honors College at Florida Atlantic University, Jupiter, FL Concentration: Chemistry Summa cum Laude G.P.A.: 3.917/4.000 Thesis: Presumptive and Confirmatory Tests for Illicit Drug Analogs

June 2007 Shoshana S. Cardin Jewish Community High School, Baltimore, MD G.P.A.: 3.83/4.00

RESEARCH AND WORK EXPERIENCE Research Assistant - Johns Hopkins University February 2014-present  Contribute to systematic reviews and methodology research for the Cochrane Eyes and Vision Group  Direct screening of literature to identify studies relevant for reviews  Extraction of relevant methodological details and results of studies for reviews

Teaching Assistant - University of Illinois at Chicago August 2011-May 2013  Hold office hours to answer students’ questions about the course material  Proctor and grade course exams 104

Research Assistant - University of Illinois at Chicago August 2011- December 2012  Utilize various chromatography techniques for the isolation and purification of natural products from plants and bacteria  Structural elucidation of natural products with mass spectrometry and nuclear magnetic resonance

BioTools Intern - Jupiter, FL August 2010-December 2010  Review literature on using infrared spectroscopy for analyzing three dimensional structure of proteins  Work with the PROTA Fourier Transform-Infrared Protein Analyzer for determining secondary structure of proteins

PUBLICATIONS AND PRESENTATIONS Journal Articles  Rouse, Benjamin; Schneider, Rebecca; Smith, Eugene. Presumptive and Confirmatory Tests using Analogs of Illicit Drugs: An Undergraduate Instrumental Methods Exercise. Journal of Chemical Education. 2014; 19:70- 72.

Poster Presentations  University of Illinois at Chicago - March 9, 2012 Chicago, Illinois College of Pharmacy Research Day Bioassay-Guided Fractionation and Dereplication of Anti-Cancer Compound from Nostoc sp. (UIC 10366) Rouse, B, Kang, HS, Zinkus, J, Swanson, S, Orjala, J

 Harriet L. Wilkes Honors College - April 15, 2011 Jupiter, FL 9th Annual Symposium for Research and Creative Projects Presumptive and Confirmatory Tests for Illicit Drug Analogs Rouse, B, Smith, E

 University of Maryland at Baltimore County - October 30, 2010 Baltimore, Maryland 13th Annual Undergraduate Research Symposium in the Chemical and Biological Sciences Qualitative Analysis of Amphetamine and GHB Analogs Rouse, B, Smith, E

105

PROFESSIONAL DEVELOPMENT Computer skills: Microsoft Office Suite (including Word, Excel, and PowerPoint) Statistical software: Stata, SAS, and R Packages

106